Elsevier

Measurement

Volume 86, May 2016, Pages 293-300
Measurement

Empirical investigation of environmental characteristic of 3-D additive manufacturing process based on slice thickness and part orientation

https://doi.org/10.1016/j.measurement.2016.03.006Get rights and content

Abstract

The significant amount of research has been done in improving the mechanical properties (compressive strength), dimensional accuracy (length, height and width), and build time of the components manufactured from the additive manufacturing process. In contrast to this, the research in the optimization of environmental characteristic i.e. energy consumption for the additive manufacturing processes such as selective laser sintering (SLS), and selective laser melting (SLM) needs significant attention. These processes intakes the significant portion of input laser energy for driving the laser system, heating system and other machine components. With world moving towards globalization of additive manufacturing processes, the optimization of laser energy consumption thus become a necessity from productivity and as well as an environmental perspective. Therefore, the present work performs the empirical investigation by proposing the optimization framework in modelling of laser energy consumption of the SLS process. The experimental procedure involves the computation of energy consumption by measuring the total area of sintering. The optimization framework when applied on the experimental data generates the functional expression for laser energy consumption which suggests that the slice thickness is a vital parameter in optimizing it. The implications arising from the study is discussed.

Introduction

The additive manufacturing processes such as selective laser sintering (SLS) and selective laser melting (SLM) is gaining considerable attention and popularity because it uses the laser energy to selectively fuse the powder into the complex shaped objected as designed using the CAD software [1], [2], [3]. The difference between SLS and SLM is that the latter involves complete melting of powder whereas the former is based on phenomenon of partial melting [4]. Extensive studies have been done in the past that focus on improving the improving the mechanical properties (compressive strength), dimensional accuracy (length, height and width), and build time of the parts manufactured from SLS process by intelligent selection of the values of input process parameters such as laser power, scan speed and scan spacing [5], [6], [7], [8], [9], [10], [11], [12]. The same notion was also stated in the work done by Garg et al. [13] on survey of empirical modelling of additive manufacturing processes. Paul and Anand [14] in his work explicitly mentioned that the SLS is extensive energy consuming process and when deployed for mass production, the inefficiency increases at higher rate resulting in increase in production cost and causes environmental problems.

The optimization of energy consumption and reducing the production cost simultaneously have become top priority for government across globe in view of rising burden of climate change. The industry however lately observe the necessity of promoting cleaner production by deploying energy managers whose sole task is to monitor the energy consumption process during the process [15]. To drive industries towards cleaner production, the government have introduced the carbon tax and imposed fines [15]. There were studies conducted to develop the models for measuring the energy consumption in the additive manufacturing processes [16], [17], [18], [19], [20]. The major component of the energy is used in driving the laser systems (Fig. 1) which exhibit higher dependence on the part properties (geometry and material), machine specifications, part orientation and the slice thickness of the SLS process [14].

Mognol et al. [17] and Niino et al. [18] evaluated the percentage of fraction of the laser energy to the total energy consumption and found the relative contributions of 66% and 1% on the two different machines (EOS EOSINT M250 Xtended and Semplice, ASPECT) respectively. The difference is attributed to the size of build platform. Smaller the size of the platform, lesser energy required for heating powder bed and moving the build platform. There were studies conducted describing the effect of input process parameters such as laser power, scan spacing, scan speed on the layers development in the SLS process by formulation of 1-D, 2-D and 3-D thermal models [21], [22], [23], [24]. The functional expressions for the (a) laser power and the inputs such as laser beam diameter, laser speed (b) laser power and the surface properties were developed [25], [26]. The evaluation of life cycle energy utilization was used to study the environmental implications from the SLS process. Fuh et al. [27] used Beer–lambert law to develop the relationship between laser power and cure depth of the laser curing process.

The past studies summarizes that the laser energy contribution to the total energy consumed during the SLS process is influenced by the type of machine used, the part geometry and other factors based on the slice thickness and part orientation. Thus, the formulation of 3-D dimension models considering the two inputs needs thorough understanding of mechanism of the SLS process. SLS process is complex in nature by occurrence of multiple phenomenon based on the heating and cooling parts and transmission and absorption of energy [14]. On the other hand, the input parameters such as the slice thickness and part orientation influencing the laser energy consumption add complexity to the process. To the best of authors’ knowledge, the limited applications of optimization algorithms in studying the energy consumption based on the slice thickness and part orientation is reported. One optimization algorithm on genetic programming (GP) [28] can be applied for formulating the functional expression between the laser energy consumption and the two inputs (slice thickness and part orientation). The potential advantage of using GP is that it uses the minimal information (only data) about the nature of process and can provide an explicit and generalized relationship for the input–output parameters [29], [30], [31].

Therefore, in this work, an optimization framework based on GP is applied to derive the function relation of the energy consumption with respect to the slice thickness and part orientation of the SLS fabricated prototype. The procedure of the modelling the given energy consumption of the SLS process is shown in Fig. 2. The energy consumption is evaluated first by experiments where the total area of sintering (TAS) is determined for every slice in the designed part. The data collected from the experiments is further then input in the optimization framework of GP for processing. The objective function used in the optimization framework of GP is based on the difference between the absolute of difference between actual and predicted values from the GP model. In this work, the framework uses the structural risk minimization principle (SRM). The formulated GP based energy consumption model is evaluated statistically and the amount of influence of the input parameters is further determined based on the sensitivity approach. The model formulated and the information mined from the statistical analysis of it is useful for the manufacturing experts for the effective monitoring of the additive manufacturing process resulting in lower energy consumption and the higher environmental performance.

Section snippets

Experimental SLS process and data collection

The experimental details and assumptions considered in this work is referred from the study conducted for evaluation of TAS and laser energy consumption by Nancharaiah et al. [32]. The settings for the machine is kept the same. In this work, the absorptivity of the laser power system, laser power, beam radius and scan speed of 0.95, 70.00 W, 17.50 um and 1 m/s respectively [6], [13]. Procedure for the evaluation of energy consumption involves the part to be modelled build in CAD and then the file

Optimization framework of GP

Genetic Programming (GP) is (Fig. 4) an evolutionary approach that mimics the process of biological evolution [28]. The mathematical models in GP are laid on symbolic regression – a type of analysis that search the space of mathematical expressions to find the best-fit model of a given dataset. Usually, these models or programs are represented by tree structures [31]. The general outline of the algorithm involved can be explained as below:

  • 1.

    The first step is where the algorithm creates a random

Statistical analysis of the GP based laser energy consumption model

This section performs the statistical analysis of the best GP based laser energy consumption model based on the following metrics:Coefficient of determination(R2)=i=1n(Ai-Ai)(Mi-Mi)i=1n(Ai-Ai)2i=1n(Mi-Mi)22Root mean square error(RMSE)=i=1N|Mi-Ai|2NMultiobjective error(MO)=MAPE+RMSER2Relative error(%)=|Mi-Ai|Ai×100where Mi is the value predicted by a model, and Yi is the actual value of the output.

Table 2 shows the values of error metrics (R2, RMSE, MAPE and MO) of the GP model on the

Dominant input parameters for the laser energy consumption of the SLS process

In this section, the sensitivity analysis is performed on the best GP model for finding the dominant parameter among the two inputs (slice thickness and the part orientation). The sensitivity analysis is done by finding the difference between the maximum and minimum from the main effect relationships between the laser energy consumption and the two inputs. The main effects are calculated by varying each input from its mean value while keeping the other input at its mean value. The values for

Conclusions

The present work addresses the need of evaluation of environmental characteristic (energy consumption) in additive manufacturing processes such as SLS. The literature in this context was studied and the motivation of finding the functional relationship for laser energy consumption based on the optimization framework is underlined. The novelty of the work lies in the proposition of optimization framework by introducing the SRM principle for generating the laser energy consumption model. The

References (32)

  • G.B.M. Cervera et al.

    Numerical prediction of temperature and density distributions in selective laser sintering processes

    Rapid Prototyping J.

    (1999)
  • J.C. Nelson

    Model of the selective laser sintering of bisphenol – a polycarbonate

    Ind. Eng. Chem. Res.

    (1993)
  • A.E. Tontowi et al.

    Density prediction of crystalline polymer sintered components at various powder bed temperatures

    Rapid Prototyping J.

    (2001)
  • A. Singh et al.

    DOE based three-dimensional finite element analysis for predicting density of a laser-sintered component

    Rapid Prototyping J.

    (2010)
  • X. Shen

    Density prediction of selective laser sintering components based on artificial neural network

  • V. Beal

    Statistical evaluation of laser energy density effect on mechanical properties of polyamide components manufactured by selective laser sintering

    J. Appl. Polym. Sci.

    (2009)
  • Cited by (31)

    • Deep learning-driven particle swarm optimisation for additive manufacturing energy optimisation

      2020, Journal of Cleaner Production
      Citation Excerpt :

      These features also play into the factors, that impact AM energy consumption, which are decided before starting the producing process by part designers and process operators (Qin et al., 2018; Ding et al., 2018). Modelling the design-relevant data of AM systems to predict energy consumption, and then reducing it by optimising the designs and decisions of part designers and process operators have become a crucial research topic for improving AM systems (Baumers et al., 2011b; Panda et al., 2016; Watson and Taminger, 2018). Among the various optimisation algorithms, particle swarm optimisation (PSO) is a powerful algorithm that is able to solve nonlinear multi-objective problems so as to help relevant professionals for decision-making (Bai, 2010; Chen and Huang, 2017; Moradi and Abedini, 2012).

    • A Comprehensive Approach for the Clustering of Similar-Performance Cells for the Design of a Lithium-Ion Battery Module for Electric Vehicles

      2019, Engineering
      Citation Excerpt :

      Future work can focus on conducting large-scale testing on cells in order to design a larger battery module, as well as on performing experimental verification on the performance of probabilistic methods [43,44], extreme machine learning methods [45,46], and artificial-intelligence-based methods [47–50].

    • Probability distribution pattern analysis and its application in the Acute Hypotensive Episodes prediction

      2017, Measurement: Journal of the International Measurement Confederation
    • A hybrid computational intelligence framework in modelling of coal-oil agglomeration phenomenon

      2017, Applied Soft Computing Journal
      Citation Excerpt :

      Choosing the appropriate training data set is important for faster and good learning capability of the GP approach. Therefore, based on the understanding of preliminary studies [30], the authors have applied 5-fold cross-validation algorithm for the generation of random five training (32 training samples) and corresponding test data (10 data samples) sets. The training data set is fed into the framework of GP in formulation of OMR (%) model.

    View all citing articles on Scopus
    View full text